jo ur nal homepage: www.elsev ier .com/ locate /e ja
heat yield response to spatially variable nitrogen fertilizer inediterranean environment
runo Bassoa,∗, Davide Cammaranob, Costanza Fiorentinoc, Joe T. Ritchiea
Department of Geological Science and Kellogg Biological Station, Michigan State University, 288 Farm Lane, East Lansing, MI 48823, USADepartment of Agricultural & Biological Engineering, University of Florida, Gainesville, FL 32611, USASchool of Agriculture, Forestry, Food, and Environmental Sciences, University of Basilicata, Viale Ateneo Lucano 10, 85100 Potenza, Italy
r t i c l e i n f o
rticle history:eceived 25 May 2013eceived in revised form 26 June 2013ccepted 29 June 2013
eywords:ariable rate nitrogenheat spatial and temporal yield variabilityanagement zone
ainfall distributionrop model
a b s t r a c t
Farmers obtain high yield when proper crop management is matched with favourable weather. Nitro-gen (N) fertilization is an important agronomic management practice because it affects profitability andthe environment. In rainfed environments, farmers generally apply uniform rates of N without takinginto account the spatial variability of soil available water or nutrient availability. Uniform application offertilizer can lead to over or under-fertilization, decreasing the efficiency of the fertilizer use. The objec-tive of this study was to evaluate the impact of variable rate nitrogen fertilizer application on spatialand temporal patterns of wheat grain yield. The study was conducted during the 2008/2009 and 2009/10growing seasons in a 12 ha field near Foggia, Italy. The crop planted each year was durum wheat (Triticumdurum, Desf.) cultivar Duilio. The field was subdivided into two management zones High (H), and Aver-age (A). Three N rates were identified using a crop model tested on the same field during a previous
−1 −1
growing season. The N rates were: low N (T1: 30 kg N ha ), average N (T2: 70 kg N ha ), and high N (T3:90 kg N ha−1). The ANOVA test showed that there were no effects of the N levels for the first growingseason for the H and A zone. For the 2009/10 growing season with higher rainfall there was a significantdifference in grain yield for the A zone (2955 kg ha−1), but not in the H zone (3970 kg ha−1). This studydemonstrates the optimal amount of N for a given management zone is not fixed but varies with therainfall amount and distribution during the fallow and growing season.
. Introduction
Nitrogen (N) fertilization is one of the most important agro-omic management practices, yet the amount and timing of Nemains a management challenge. Crops growing with N deficiencyose greenness, they are usually smaller with less biomass, and haveeduced photosynthetic capacity resulting in poor yield and lowrotein content. It has been estimated that more than 50% of theuman population relies on N fertilizers for food production (FAO,008). Their global demand has increased of about 7.3 million tonsf N per year (IFAI, 2008). Proper N management is important forainfed agriculture (annual rainfall between 200 and 600 mm). Inainfed environments, if N is available, crops can use much of thevailable soil water prior to anthesis, leaving the soil dry during therain filling with resulting low yield and poor grain quality without
dequate rainfall during the grain filling period. Soil water supplyeeds to be adequate before anthesis to develop a good canopy andrain set around anthesis, and during grain filling. Although wheat
yield responds positively to soil water content, response will varybased on soil properties and rainfall distribution within the sea-son (Norwood, 2000; Kirkegaard et al., 2001; Nielsen et al., 2002;Sadras, 2002; Anderson, 2010). In rainfed winter wheat areas, farm-ers generally apply uniform rates of N, with approximately 30%at sowing and 70% at the time of early stem elongation withouttaking into consideration the spatial variability of soil or rainfalldistribution (Saseendran et al., 2004). This can lead to over- orunder-fertilization, decreasing the efficiency of the fertilizer use(Mulla et al., 1992). Moreover, excessive N application will result inpotential environmental impact due to nitrate leaching, ammoniavolatilization, nitrous oxide emissions and soil acidification (Chenet al., 2008a,b; Li et al., 2008). In Germany, Ehlert et al., 2004 foundthat in variable rate N application, N fertilizer could be reduced by12% without affecting wheat yield. Farmers can increase the effi-ciency of N fertilization, maximizing crop N uptake, and minimizingN losses by taking into account the spatial and temporal N needof the crop by dividing the field into homogenous zones of simi-
lar behaviour using geospatial technologies (Basso et al., 2011a,b).Robertson et al. (2005, 2007) found that the benefits of dividing thefields into homogenous zones can be achieved if the field zones areconsistent in yield performance. Basso et al. (2012) defined three
patially and temporally stable zones by integrating spatial mea-urements of soil physical properties, five years of remotely sensedata and yield monitoring from 2006 to 2011 in the Mediterraneannvironment of Southern Italy. The interaction between the spatialariability of soil properties and the rainfall distribution affectedhe spatial and temporal variability of grain yield. Each zone had aifferent response to fallow and growing season rainfall. In the highielding stable zones with high spatial and temporal stability, a highorrelation between fallow rain and grain yield was observed, buthis occurred only if the growing season rainfall was adequate. Theverage yield stable zones did not respond to fallow rain because of
shallow soil rooting depth and grain yield was more responsivef the growing season rainfall was not excessive.
The interaction between rainfall, position in the landscape andoil properties is important when selecting the amount of N to applyt side-dressing to improve N use efficiency.
The hypothesis of this study was that wheat yield responseo nitrogen fertilizer varies over space and time in predefinedomogenous zones. The objective of this study was to evaluate the
mpact of variable rate nitrogen fertilizer application on spatial andemporal patterns of wheat grain yield.
. Materials and methods
.1. Site description and agronomic management
The study was carried out on a 12 ha field located in Foggia,taly (41◦27′47′ ′ N, 15◦30′24′ ′ E; 80 elev.) during two growing sea-ons 2008/2009 and 2009/10. The soil is a deep silty-clay Vertisolf alluvial origin, classified as fine mesic Typic, Chromoxerert (Soilurvey Staff, 1999). The crop planted each year was durum wheatTriticum durum, Desf.) cultivar Duilio. For each growing season theeedbed was prepared the first week of September with a mini-um tillage (chisel plough) at a depth of 20 cm. The sowing was
arried out in both years the first week of December at a depthf 5 cm with 17 cm distance between the rows and with a densityf 400 plants m−2. The homogenous zones used in this study areescribed in Basso et al. (2012) and are High yield zone (H), Aver-ge yield zone (A). The zones were selected adopting a modifiedersion of the methodology first reported by (Blackmore, 2000).heat yield spatial variability was analyzed, by calculating the rel-
tive percentage difference of yield crop from the average yieldevel obtained within the field at each point mapped. Overlay-ng the single map of the relative percentage difference createdhe final map of the zones with different yield levels. The tempo-al variability of yield patterns, expressed as degree of stability,as calculated as temporal variance (yield value recorded at eachoint mapped minus the field mean) according to the methodroposed by Blackmore et al. (2003). By overlaying the map of rela-ive percentage difference with the one of temporal variance, highnd stable yield zones (H) were identified. Temporally stable andverage yield were defined as A zones. The parts of the field charac-erized by unstable yield were classified as unstable (U) and wereot considered in the analysis.
The amount of N fertilizer to apply in each zone was derivedy using the SALUS crop simulation model over thirty years witheather and soil properties from the study site and following the
ame approach described by (Basso et al., 2010, 2011a,b). The modelas tested using an independent dataset of grain yield from the
rowing season 2007/08 (Basso et al., 2010). The three N rates iden-ified by the model were: low N rate (T1: 30 kg N ha−1), average
rate (T2: 70 kg N ha−1), and high N rate (T3: 90 kg N ha−1). Theost common N application in the study area is about 90 kg N ha−1
ith 30% given at sowing and 70% at early stem elongation. Thereatments were randomly distributed within each zone (Fig. 2).
nomy 51 (2013) 65– 70
Fertilization consisted in two N split applications, one at sowingwith 30 kg N ha−1 (representing 30% of the application) and theremaining 70% as a variable dose. The type of N fertilizer appliedwas ENTEC 25-15. The dates for the second N applications were 16March 2009 and 17 March 2010. The field was divided into 27 regu-lar and adjacent subareas (30 m × 10 m). Three plots were excludedfrom further analysis because in each plot there were the samenumber of pixels of A and H (Fig. 1b).
2.2. Field measurements
Grain yield was georeferenced using a yield monitor system(grain mass flow and moisture sensors). Site coordinates for eachyield measurement were determined with a differentially cor-rected GPS (OMNISTAR signal) Trimble 132 receiver with 1 cmaccuracy. The SMS software version 3.0TM (AgLeaderTM Technol-ogy, Inc.) was used to read the raw yield data (expressed at 13.5%dry matter). Yield data were then processed to eliminate outliervalues lower than 500 kg ha−1 and greater than 7000 kg ha−1. Theyield maps were obtained by plotting the yield data, using linearinterpolation, at the nodes of a regular grid of 5 m spatial resolution.The yield maps were georeferenced and recorded in UTM WGS 84zone 33 N.
2.3. Statistical analysis
The N plots were large enough (about 300 m2) to be treated asindependent replications for statistical analysis using the conven-tional analysis of variance (ANOVA). The ANOVA was performed onthe mean yield per plot values during both study years (2008/09 and2009/2010) with a reference significance value of 0.05. The analy-sis was performed separately for the two homogenous zones. TheANOVA test was used to evaluate the influence of the homogenouszones on the relationship between N applied and grain yield. In theA zone, there were 14 plots of which 6 plots had T1, 5 plots T2, and3 plots T3. In the H zone there were 2 plots with T1, 2 plots withT2, and 6 plots with T3. Since the T1 and T2 plots’ distribution in H,the Kruskal–Wallis non-parametric test (Kruskal and Wallis, 1952)was used for testing if the plots originate from the same distribu-tion. The model was evaluated using the root mean square error(RMSE):
RMSE =[
1n
n∑i=1
(Qi − Pi)2
]1/2
(1)
where Qi is the observed value, Pi is the simulated value, and n isthe number of pairs of measured and simulated values. In addi-tion, the d-index or index of agreement (Willmott, 1982) which isa descriptive measure was calculated as follows:
d − index = 1 −[ ∑n
i=1(Pi − Oi)2∑n
i=1(|P ′i| + |O′
i|)2
](2)
where n is the number of observed values, Oi is the observedvalues, Pi the simulated values, P′
i = Pi − Ôı and O′ı = O′
ı − Ô′ı (Ôı is
the average of the observations). The d-index values range between0 (not good agreement) and 1 (good agreement).
For each of the two management zones and for the whole fieldthe variability of grain yield was quantified using the coefficientof variation calculated for the 2008/09 and 2009/10 growing sea-sons. This variability was calculated using the average yields andstandard deviations of the whole field, the H zone and the A zone
as follows:
CV [%] = �
�× 100 (3)
B. Basso et al. / Europ. J. Agronomy 51 (2013) 65– 70 67
Fig. 1. (a) Spatial distribution of the homogeneous zones (adapted from Basso et al., 2012) and (b) plots of homogeneous zones, the black plots were excluded from theanalysis because they had about the same number of pixels of Average (A) and High (H) zone.
Fig. 2. Spatial distribution of the randomly determined N treatments for the growing seasons 2008/09 and 2009/10.
68 B. Basso et al. / Europ. J. Agro
Table 1Measured and simulated yield (kg ha−1), standard deviation, root mean square error,and d-index for the whole field, the Average zone (A), and the High zone (H) for thegrowing season 2007–2008.
here � is the standard deviation of the yield (kg ha−1), and � ishe mean of the yield (kg ha−1). ANOVA analysis was performed in
atlab software (MATLAB®, R2011b).
. Results
Results of the model testing are shown in Table 1. The grainield data are from an independent dataset from the growing sea-on 2007/08. The RMSE was 163.5 kg ha−1 and the d-index 0.80Table 1). Fallow and growing season rainfall for the growing sea-ons 2008/09 and 2009/10 are shown in Fig. 3. Fallow rainfallSeptember–November) was 172 mm for 2008/09 and 207 mmor 2009/10. The total growing season rainfall (December–May)as 564 mm and 360 mm for the growing season 2008/09 and
009/10, respectively (Fig. 3). The partition of the growing sea-on rain into monthly values were higher in the 2008/2009 seasonxcept February and May when were higher for the 2009/10 grow-ng season.
The spatial variability of grain yield for the two growing sea-ons is shown in Fig. 4a–b. Overall, wheat yields varied between00 and 5000 kg ha−1 with an average yield of 3000 kg ha−1 and500 kg ha−1 for the growing season 2008/09 and 2009/10, respec-ively. Grain yield within each N treatment varied spatially withinhe field and within each management zone. For the first grow-ng season the 30 N and 70 N treatments had yield values rangingetween 2150 and 3200 kg ha−1 and 90 N treatments had yield val-es ranging between 2500 and 3200 kg ha−1 (Fig. 4c). In the secondeason grain yields within each plot were higher than the first sea-on, with 30 N yields ranging between 2400 and 3100 kg ha−1, 70 Netween 2650 and 3000 kg ha−1, and 90 N ranging between 2650nd 3400 kg ha−1 (Fig. 4d). Some of the N plots had no increase
r decrease in yield between the two seasons. For example, 30 N,0 N and 90 N located on the left-mid portion of the field had sim-
lar yields for both years, while the same plots located in the rightottom-mid portion of the field had an increase in yield from the
ig. 3. Fallow (September–November) rainfall, monthly rainfall, and rainfalletween the second N application for the growing season 2008/09 and 2009/10.
nomy 51 (2013) 65– 70
first to the second growing season (Fig. 4c–d). In the H zone the30 N and 90 N yields for the second growing season had the sameyield range (Fig. 4d) demonstrating that 30 N was sufficient for thiszone.
Results of the ANOVA analysis for the first growing seasonshowed non-significant effect of the N levels on grain yield atp < 0.05, but in the second season there was a significant N responseon the yield. The ANOVA test was used for each managementzone analyzed separately. The A zone in the first growing sea-son had no N treatment effect on grain yield [F(2,12) = 0.23,p-value = 0.798] while there was a significant N effect on grainyield for the second season [F(2,12) = 6.98, p-value = 0.01). In theH zone the Kruskal–Wallis non-parametric test showed no sig-nificant effects of N treatments on grain yield for both growingseasons [F(2,7) = 0.52, p-value = 0.61 for 2008/09; F(2,7) = 1.18, p-value = 0.36 for 2009/10].
The CV for the first growing season was 22% for the whole field,14% for the H zone and 16% for the A zone. In the second year, it waslower, with 16, 10, 12% for the whole field, H, and A, respectively.
The grain yield measured for each N treatment, each manage-ment zone, and for the two growing seasons are shown in Fig. 5.Overall, there was a large variability in yield response between thetreatments within each zone and between the years. In the firstyear the grain yield in the A zone ranged between 12177 kg ha−1
and 2566 kg ha−1 for T1, between 2335 kg ha−1 and 2863 kg ha−1 forT2 and between 2587 kg ha−1 and 2656 kg ha− for T3. For the sameyear the H zone grain yield was 3167 kg ha−for T1, 2800 kg ha−1 and3000 kg ha−1 for T2, and between 2561 kg ha−1 and 3098 kg ha−forT3. In the first growing season the maximum yields for the H werehigher for the T1 lower N rate of T3 than for the T3 but the differ-ences were not statistically different (Table 1). In the second seasonT3 produced higher yields than the previous growing season withH having the higher yields. T1 treatment had higher yield for the Hzone but they were not statistically different (Table 1). In both zonesthe T2 treatment had little variation in the range of the measuredyield (Fig. 5).
4. Discussion
Results of this study complement a previous study for the samefield (Basso et al., 2012). The CV% values and spatial yield variabil-ity (Fig. 4) of the present study agree with the results of Basso et al.(2012) in Table 1. The earlier study highlighted the importance offallow and growing season rainfall and stored soil water content atthe time of second N application on spatial and temporal variabilityof wheat yields. It also demonstrated the interaction between rain-fall and spatial distribution of soil properties within each uniformmanagement zone.
The lack of any significant correlation between yield and the Ntreatments for each of the uniform zone (Fig. 5) can be explainedby the patterns of the fallow and growing season rainfall and forthe spatial soil properties of each uniform zone analyzed in Bassoet al. (2012). In wet years the whole field requires less N becauseof water logging in the low elevation zones of the field (Bassoet al., 2012) negatively influences the plant population stand. Apoor wheat stand will take put a lower amount of N. The 2008/09growing season rainfall was 204 mm higher than the 2009/10 sea-son; therefore the benefits of stored water prior to sowing (fallowrain) was less evident. The H zone has a deeper soil profile for rootgrowth and uptake (Basso et al., 2012). This allows more rainfallto be stored during fallow and usually has a positive influence
on grain yield. Information in Figs. 4d–c and 5 shows that theyield in the T3 and T1 treatments in the lower right portion ofthe field are not different. This suggests that in the lower por-tion of the field the lower amount of N (T1) would be adequate
B. Basso et al. / Europ. J. Agronomy 51 (2013) 65– 70 69
F 9 andg grain
iiTrdaeNt
ig. 4. (a) Spatial distribution of grain yield (kg ha−1) for the growing season 2008/0rain yield for each N plot for the growing season 2008/09 growing season, and (d)
n years where the rainfall up to the second N application is sim-lar to the ones observed in this study for both growing season.hus the H zone will have little advantage from high fertilizationates because of the high content in organic matter (>2%) and theeeper exploitable root zone (Basso et al., 2012). Overall, this result
grees with the findings of Basso et al. (2010) that for this typenvironment, lower N fertilization rate is sufficient to meet crop
demands, but this amount is dictated by the spatial soil proper-ies.
(b) Spatial distribution of grain yield (kg ha−1) for the growing season 2009/10; (c)yield for each N plot for the growing season the 2009/10 growing season.
The lack of response of yield to N treatments for the whole fieldin 2008/09 is also related to the high amount of rainfall. During themonth of December, the 116 mm of rainfall caused water loggingin the low elevation zones of the field; moreover 77 mm rain fellbetween the second N application (16 March) and 25 days later
(April 10) (Fig. 3). The high rainfall in a short time could causeN leaching, especially for the T2 and T3 treatments. Basso et al.(2012) found that in the A zone there is a negative correlationbetween grain yield and growing season rainfall because of water
70 B. Basso et al. / Europ. J. AgroY
ield
(kg
ha
-1)
2000
2200
2400
2600
2800
3000
3200
3400
A 2009
H 2009
A 2010
H 2010
30T1
N Rates (kg N ha-1
)
60T2
90T3
Fg
lhisrtafwdtbr
mohobedficctpetma
nratufstwTd
ig. 5. Measured grain yield for the all treatment for the 2008/2009 (closed trian-les) and 2009/2010 (open triangles) growing seasons.
ogging in this depression area (Basso et al., 2012). In the H zoneigh N fertilization will not always result in higher yields as shown
n Fig. 4d for the mid-portion of the field where T1 treatmentshow the same yield levels as T2. The July–September fallow periodainfall was higher in 2009/10 than the previous growing season,hus the benefits of stored water prior to sowing was beneficial forpparently causing low yield variability within the field as shownrom Fig. 5. Our findings agree with those of Sadras et al. (2012)ho demonstrated that July–September fallow (approximately 60ays before sowing) water storage had more benefits on grain yieldhan July–December fallow (time between the previous harvest andefore the next sowing) water storage and the benefits of fallowain declined as within season rainfall increased.
The A zone had a response to N for the 2009/10 growing seasonainly because the entire zone did not respond to fallow rainfall
wing to its soil physical properties (Table 1 and Fig. 4d). The soilas higher clay content in the upper 50 cm over a compacted layerf soil and stones because the central transect was an old creeked (Basso et al., 2012). The upper right area soil profile had a highlectrical resistivity with high silt and coarse sand fraction and aescending slope in the direction of the upper right corner of theeld shown Basso et al. (2012) in Fig. 5 of that paper. These factorsontribute to the lower soil water storage in the shallow profile andontribute to a response to different N levels (Basso et al., 2012) dueo different amount of water available to plants and root exploitablerofile which will determine the amount of N uptake. The knowl-dge of spatial variability of plant extractable soil water duringhe off-season, at planting and at side dressing is crucial to opti-
ize nitrogen fertilizer. Such knowledge can be obtained through validated simulation approach using long-term weather records.
In conclusion, this study demonstrates how the amount of Needed for spatially variable fields is not fixed but varies with theainfall amount and distribution of rainfall within the fallow periodnd growing season as well as the stored soil water at the time ofhe spring N fertilizer application. Within each defined somewhatniform management zone there is an appropriate amount of Nor optimum grain yield within the confines of the uncertainty ofpatial and temporal variability. The study showed experimentally
hat N fertilizer can be reduced without compromising yield, andith the consequent advantage of reducing nitrous oxide emission.
he simulation of the soil water balance in each zone and the rainfallistribution can be used to make tactical decisions regarding the N
nomy 51 (2013) 65– 70
fertilizer management, thus providing the opportunity to increaseprofitability and decrease N losses as demonstrated in Basso et al.(2011a,b, 2012). Thus, this type of management strategy can beadopted in fields where spatial variability exists.
References
Anderson, W.K., 2010. Closing the gap between actual and potential yield of rain-fed wheat. The impacts of environment, management and cultivar. Field CropResearch 116, 14–22.
Basso, B., Fiorentino, C., Cammarano, D., Cafiero, G., Dardanelli, J., 2012. Analysis ofrainfall distribution on spatial and temporal patterns of wheat yield in Mediter-ranean environment. European Journal of Agronomy 41, 52–65.
Basso, B., Cammarano, D., Troccoli, A., Chen, D., Ritchie, J.T., 2010. Long-term wheatresponse to nitrogen in a rainfed Mediterranean environment: field data andsimulation analysis. European Journal of Agronomy 33, 132–138.
Basso, B., Sartori, L., Bertocco, M., Cammarano, D., Grace, P.R., 2011a. Economic andenvironmental evaluation of site-specific tillage in a maize crop in NE Italy.European Journal of Agronomy 35, 83–92.
Basso, B., Ritchie, J.T., Cammarano, D., Sartori, L., 2011b. A strategic and tactical man-agement approach to select optimal N fertilizer rates for wheat in a spatiallyvariable field. European Journal of Agronomy 35, 215–222.
Blackmore, S., Godwin, R.J., Fountas, S., 2003. The analysis of spatial and temporaltrends in yield map data over six years. Biosystems Engineering 84 (4), 455–466.
Blackmore, B.S., 2000. The interpretation of trends from multiple yield maps. Com-puters and Electronics in Agriculture 26, 37–51.
Chen, D., Suter, H., Islam, A., Edis, R., Freney, J.R., Walker, C.N., 2008a. Prospects ofimproving efficiency of fertiliser nitrogen in Australian agriculture: a review ofenhanced efficiency fertilisers. Australian Journal of Soil Research 46, 289–301.
Chen, D., Li, Y., Grace, P., Mosier, A.R., 2008b. N2O emissions from agricultural lands:a synthesis of simulation approaches. Plant and Soil 309, 169–189.
Ehlert, D., Schmerler, J., Voelker, U., 2004. Variable rate nitrogen fertilisation ofwinter wheat based on a crop density sensor. Precision Agriculture 5, 263–273.
Food and Agricultural Organization, 2008. Current world fertilizer trends and out-look to 2011/12. FAO, Rome, Italy.
International Fertilizer Association Industry, 2008. 76th IFA Annual Conference ,Vienna, Austria, 19–21 May.
Kirkegaard, J.A., Howe, G.N., Pitson, G., 2001. Agronomic interactions betweendrought and crop sequence “Science and technology: delivering results foragriculture”. In: Proceedings of the 10th Australian Agronomy Conference, Aus-tralian Society of Agronomy, Hobart, Australia www.regional.org.au/au/asa/2001/4/c/kirkegaard1.htm
Kruskal, W.H., Wallis, W.A., 1952. Use of ranks in one-criterion variance analysis.Journal of the American Statistical Association 47 (260), 583–662.
Li, Y., Chen, D., Baker-Reid, F., Eckard, R., 2008. Simulation of N2O emissions fromrain-fed wheat and the impact of climate variation in southeastern Australia.Plant and Soil 309, 239–251.
2011. Matlab® version 1.6.0 17. The MathWorks Inc., Natick, Massachusetts, USA.Mulla, D.J., Bhatti, A.U., Hammond, M.W., Benson, J.A., 1992. A comparison of win-
ter wheat yield and quality under uniform versus spatially variable fertilisermanagement. Agriculture, Ecosystems & Environment 38, 301–311.
Nielsen, D.C., Vigil, M.F., Anderson, R.L., Bowman, R.A., Benjamin, J.G., Halvorson,A.D., 2002. Cropping system influence on planting water content and yield ofwinter wheat. Agronomy Journal 94, 962–967.
Norwood, C.A., 2000. Dryland winter wheat as affected by previous crops. AgronomyJournal 92, 121–127.
Robertson, M., Isbister, B., Maling, I., Oliver, Y., Wong, M., Adams, M., Bowden,B., Tozer, P., 2005. Managing spatial and seasonal variability within field canimprove the profitability of WA grain production. In: Turner, Neil, Acuna, Tina(Eds.), Groundbreaking Stuff. Proceedings of the 13th Australian Agronomy Con-ference, 10–4 September 2006 Perth, Western Australia.
Robertson, M., Isbister, M., Maling, I., Oliver, Y., Wong, M., Adams, M., Bowden, B.,Tozer, P., 2007. Opportunities and constraints for managing within-field spatialvariability in Western Australian grain production. Field Crops Research 104,60–67.
Sadras, V.O., 2002. Interaction between rainfall and nitrogen fertilisation of wheatin environments prone to terminal drought: economic and environmental riskanalysis. Field Crops Research 77, 201–215.
Sadras, V.O., Lawson, C., Hooper, P., McDonald, G., 2012. Contribution of sum-mer rainfall and nitrogen to the yield and water use efficiency of wheatin Mediterranean-type environments of South Australia. European Journal ofAgronomy 36, 41–54.
Saseendran, S.A., Nielsen, D.C., Ma, L., Ahuja, L.R., Halvorson, A.D., 2004. Modellingnitrogen management effects on winter wheat production using RZWQM and
